I'll be honest, if our customers were not as stubborn as they are about hosting their own software then I would never do anything else but host it myself either!
We host a lot of sites on our custom CMS and it makes life so much easier to role bug fixes and db updates when there is only 1 code base to update. Also you don't have issues where people are on older version and need to update through versions
We try push all our customers into this model with those exact reasons.
It's a dangerous road when you value your own operations over customer needs. I'd say invest in the infrastructure needed to one-click deploy outside your data center as well.
This may be a solution further down the road but our code at the minute changes literally daily and it would be counter productive to have to test every deployment against previous versions
We actually discussed this very point in work today, I still class our product, 2 years down the line, to be in beta. We are actively working towards a version 1, if you will.
We could revisit at a later point and probably will but the time it would take to document all db changes and create a version every day would far outweigh the benefits.
Distributing to other peoples data centers comes with it's own host of issues as well, we would have to charge for a maintenance contract for 1. We would need to bring on a customer service team to support people. At the minute, bug fixes can come straight to developers via our technical director as we can deploy straight away. We would need a team to asses what their setup is, if they have everything needed installed etc etc.
Deployment to our own servers helps us and the customer a lot, we don't charge a lot for hosting, our uptime last year was over 99.97% and bug fixes and functionality are rolled out on a daily basis without interaction from the customer.
We did a massive project recently that the customer ended up deployment on their own $5 hosting company, turned out the company didn't support what we needed and we ended up coding around the road blocks. Ended up costing the customer more.
So there's a lot of extra work to do for self hosting customers. But isn't the question really whether your customers are ready to pay for that extra work?
In the specific case of my current employer I can say that we have enough budget to pay a few dollars per seat per day without blinking, and we'd choose this option even if a hosted Trello option was ten times cheaper.
There are two main constraints that need to be taken into account where I work when it comes to buying things:
Firstly, variable pricing models are a no-go. We'll happily pay well over the odds to secure abundant capacity but we need to know in advance with absolute certainty exactly how much it will cost for a full year of service.
Secondly, we deal with potentially sensitive information that could be damaging to our organisation and our government if it were in any way compromised. Because of this we need to retain control of the data we'd store in any kind of issue tracking or collaboration system.
Trello is clearly a great tool and I hope Joel et al do well with it. It's a great tool for agile teams to use, and I can only blame the bureaucracy of my organisation for preventing my small team's ability to use it.
I tried to pitch Trello to a company I know that I think could have really made good use of it, but their objection was the same, uncomfortable with not having their development tasking and customer issues stored "in house."
We host a lot of sites on our custom CMS and it makes life so much easier to role bug fixes and db updates when there is only 1 code base to update. Also you don't have issues where people are on older version and need to update through versions
We try push all our customers into this model with those exact reasons.